This document looks into the relationship between population and area by country using the R programming language. The data used for this analysis is from this wikipedia page.
The ulimate goal of this analysis is to determine countries who are over/under populated when compared against countries with similar area.
Firstly, read in the data. This is farily easy to do in wikipedia. Following this blog post helped me.
html <- "https://en.wikipedia.org/wiki/List_of_countries_and_dependencies_by_population_density"
xpath <- '//*[@id="mw-content-text"]/div/table'
global <- html %>% read_html() %>%
html_nodes(xpath = xpath) %>% html_table()
pop_raw <- global[[1]] %>% as_tibble() %>% clean_names()
head(pop_raw)## # A tibble: 6 x 9
## pos country_or_depe… area_km2 area_mi2 population density_pop_km2
## <chr> <chr> <chr> <chr> <chr> <chr>
## 1 • World (land onl… 134,940… 52,100,… 7,745,375… 57
## 2 • World (land onl… 148,940… 57,510,… 7,745,375… 52
## 3 • World (with wat… 510,072… 196,940… 7,745,375… 15
## 4 – Macau (China) 32.9 13 667,400 20,286
## 5 1 Monaco 2.02 0.78 38,300 18,960
## 6 2 Singapore 722.5 279 5,638,700 7,804
## # … with 3 more variables: density_pop_mi2 <chr>, date <chr>,
## # population_source <chr>
Looking at the first 6 rows we can see that this data is in need of a good clean. In the next section of code I remove the commas and mutate the column to their correct types.
pop <- pop_raw %>%
mutate_at(c(
"area_km2",
"area_mi2",
"population",
"density_pop_km2",
"density_pop_mi2"
),
~ str_replace_all(., ',', '') %>% as.numeric()
) %>%
mutate(date = mdy(date),
country_or_dependent_territory = str_remove(country_or_dependent_territory,'( *)\\[note .*')) %>%
filter(!str_detect(country_or_dependent_territory,"World") ) %>%
mutate_at(c("population","area_km2"),list(log = log))
head(pop)## # A tibble: 6 x 11
## pos country_or_depe… area_km2 area_mi2 population density_pop_km2
## <chr> <chr> <dbl> <dbl> <dbl> <dbl>
## 1 – Macau (China) 32.9 13 667400 20286
## 2 1 Monaco 2.02 0.78 38300 18960
## 3 2 Singapore 722. 279 5638700 7804
## 4 – Hong Kong (Chin… 1106 427 7482500 6765
## 5 – Gibraltar (UK) 6.8 2.6 33140 4874
## 6 3 Vatican City 0.44 0.17 1000 2273
## # … with 5 more variables: density_pop_mi2 <dbl>, date <date>,
## # population_source <chr>, population_log <dbl>, area_km2_log <dbl>
This looks much neater.
Now lets look at the relationship. As the below graph shows, their seems
to be a positive relationship between area_km2_log and the
population_log. I’ve transformed the predictor variable as there
appears to be non-linear associations in the data. As well as this, I’ve
transformed the response variable to address the issue of non-constant
variance.
pop %>% ggplot(aes(x = area_km2_log, y = population_log)) +
geom_point() + geom_smooth(method = "lm")
Now let’s build a model to help quanitfy this relationship.
pop_model <- lm(population_log ~ area_km2_log,pop)
summary(pop_model)##
## Call:
## lm(formula = population_log ~ area_km2_log, data = pop)
##
## Residuals:
## Min 1Q Median 3Q Max
## -7.2984 -0.6545 0.1625 0.8472 3.9283
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.83910 0.29710 23.02 <2e-16 ***
## area_km2_log 0.75677 0.02771 27.31 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.529 on 249 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.7497, Adjusted R-squared: 0.7487
## F-statistic: 745.8 on 1 and 249 DF, p-value: < 2.2e-16
Looking at the summary, it appears there is a relationship between the two variables. The p value is less than 5% so we can reject the null hypothesis. The R squared value of 0.75 informs us that the relationship is strong but there is room for improving this model.
Let’s look at the residual plots to see if model follows assumptions.
par(mfrow=c(2,2))
plot(pop_model)
As this is only for fun, I’m happy with the residual plots and model statistics and feel it is suited for the purpose of this analysis.
Let’s obtain residuals and convert them back into normal figures rather than log of population.
pop_aug <- pop_model %>% augment(pop) %>%
mutate(.resid = round(exp(.fitted)-population),
.fitted = round(exp(.fitted)),
error = round(population/.fitted,2))Looking at the below, Macau and Singapore are two of the countries who are over populated when compared against countries with similar area.
UK makes the top 20.
view_pop <- pop_aug %>% filter(population > 100000) %>% select(country_or_dependent_territory,area_km2,population,.fitted,.resid,error)
view_pop %>% arrange(desc(error)) %>% head(20) %>% kable(caption = "Over Populated Countries",format = "markdown",format.args = list(big.mark = ",",scientific =F))| country_or_dependent_territory | area_km2 | population | .fitted | .resid | error |
|---|---|---|---|---|---|
| Macau (China) | 32.9 | 667,400 | 13,133 | -654,267 | 50.82 |
| Singapore | 722.5 | 5,638,700 | 136,040 | -5,502,660 | 41.45 |
| Hong Kong (China) | 1,106.0 | 7,482,500 | 187,762 | -7,294,738 | 39.85 |
| Bangladesh | 143,998.0 | 167,533,646 | 7,479,575 | -160,054,071 | 22.40 |
| India | 3,287,240.0 | 1,354,430,150 | 79,786,890 | -1,274,643,260 | 16.98 |
| Bahrain | 778.0 | 1,543,300 | 143,877 | -1,399,423 | 10.73 |
| South Korea | 100,210.0 | 51,811,167 | 5,684,951 | -46,126,216 | 9.11 |
| Taiwan | 36,197.0 | 23,593,794 | 2,630,592 | -20,963,202 | 8.97 |
| Philippines | 300,000.0 | 108,480,840 | 13,034,936 | -95,445,904 | 8.32 |
| Japan | 377,944.0 | 126,140,000 | 15,524,504 | -110,615,496 | 8.13 |
| China | 9,640,821.0 | 1,399,866,680 | 180,118,171 | -1,219,748,509 | 7.77 |
| Pakistan | 803,940.0 | 206,580,217 | 27,484,284 | -179,095,933 | 7.52 |
| Palestine | 6,020.0 | 4,976,684 | 676,817 | -4,299,867 | 7.35 |
| Malta | 315.0 | 493,559 | 72,582 | -420,977 | 6.80 |
| Vietnam | 331,212.0 | 94,660,000 | 14,048,778 | -80,611,222 | 6.74 |
| Lebanon | 10,452.0 | 6,855,713 | 1,027,533 | -5,828,180 | 6.67 |
| Nigeria | 923,768.0 | 200,962,000 | 30,531,451 | -170,430,549 | 6.58 |
| Rwanda | 26,338.0 | 12,374,397 | 2,068,003 | -10,306,394 | 5.98 |
| United Kingdom | 242,910.0 | 66,435,600 | 11,110,440 | -55,325,160 | 5.98 |
| Netherlands | 41,526.0 | 17,356,488 | 2,918,724 | -14,437,764 | 5.95 |
Similarly, we can see French Guiana and Western Sahara are two countries who are currently under populated when compared against countries of similar area. Western sahara is no surpise as it is predominately desert and from a quick google search it appears French Guiana comprises of mostly tropical rainforests.
view_pop %>% arrange(error) %>% head(20) %>% kable(caption = "Under Populated Countries",format = "markdown",format.args = list(big.mark = ",",scientific =F))| country_or_dependent_territory | area_km2 | population | .fitted | .resid | error |
|---|---|---|---|---|---|
| French Guiana (France) | 86,504 | 244,118 | 5,086,138 | 4,842,020 | 0.05 |
| Western Sahara | 252,120 | 567,421 | 11,427,786 | 10,860,365 | 0.05 |
| Iceland | 102,775 | 360,390 | 5,794,733 | 5,434,343 | 0.06 |
| Suriname | 163,820 | 568,301 | 8,246,392 | 7,678,091 | 0.07 |
| Mongolia | 1,564,100 | 3,000,000 | 45,480,184 | 42,480,184 | 0.07 |
| Guyana | 214,999 | 782,225 | 10,130,144 | 9,347,919 | 0.08 |
| Namibia | 825,118 | 2,413,643 | 28,030,459 | 25,616,816 | 0.09 |
| Botswana | 581,730 | 2,302,878 | 21,515,747 | 19,212,869 | 0.11 |
| Mauritania | 1,030,700 | 3,984,233 | 33,170,084 | 29,185,851 | 0.12 |
| Libya | 1,770,060 | 6,470,956 | 49,943,456 | 43,472,500 | 0.13 |
| Artsakh | 11,458 | 150,932 | 1,101,534 | 950,602 | 0.14 |
| New Caledonia (France) | 18,575 | 258,958 | 1,587,756 | 1,328,798 | 0.16 |
| Gabon | 267,667 | 2,067,561 | 11,957,179 | 9,889,618 | 0.17 |
| Australia | 7,692,024 | 25,512,080 | 151,823,297 | 126,311,217 | 0.17 |
| Canada | 9,984,670 | 37,804,808 | 184,958,951 | 147,154,143 | 0.20 |
| Belize | 22,965 | 398,050 | 1,864,279 | 1,466,229 | 0.21 |
| Central African Republic | 622,436 | 4,737,423 | 22,645,674 | 17,908,251 | 0.21 |
| Vanuatu | 12,281 | 304,500 | 1,160,902 | 856,402 | 0.26 |
| Abkhazia | 8,660 | 243,206 | 891,213 | 648,007 | 0.27 |
| Kazakhstan | 2,724,900 | 18,356,900 | 69,225,992 | 50,869,092 | 0.27 |
There is still work to do to improve model. Adding the following may be good starting points.
- Average temperature
- GDP per capita
- Mountinous area